What podcasts say about Nufar Gaspar
Every statement, with the speaker, the exact quote and the moment it was said.
What Nufar Gaspar has said on podcasts
74 statements · 44 positive · 15 negative · 4 mixed · 11 neutral
- on Blind AI model evaluationPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Blind evaluation should hide model and tool identities to reduce evaluator bias.
“we'll be advised to hide the names of which models or which tools created each result because we are very biased towards our beloved tools”
Open the episode · The Best Way to Test New AI ModelsListen at 6:15
- on AI model switchingNeutralOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
A single benchmark win does not automatically justify changing the current AI setup.
“the fact that in a specific benchmark run, one tool or one model outperformed your current setup does not necessarily mean that automatically you need to make changes”
Open the episode · The Best Way to Test New AI ModelsListen at 6:31
- on Personal AI benchmarksPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Personal benchmarks should include one or two previously unsuccessful wishlist tasks.
“I would strongly recommend that you have one or two of those”
Open the episode · The Best Way to Test New AI ModelsListen at 13:16
- on Personal AI benchmark use casesPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
A personal AI benchmark should contain roughly four to six use cases.
“My recommendation would be like four to six type of such use cases”
Open the episode · The Best Way to Test New AI ModelsListen at 13:34
- on AI model candidatesPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Evaluations should compare the existing model with no more than four candidates.
“I would recommend to run the existing model versus a new one or up to four different candidates”
Open the episode · The Best Way to Test New AI ModelsListen at 15:11
- on AI benchmark executionPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Each benchmark request should run in a fresh chat to avoid context effects.
“we want to run every request in a fresh chat such that the context window will not send us on a tailspin”
Open the episode · The Best Way to Test New AI ModelsListen at 15:23
- on High-stakes AI benchmarksPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
High-stakes benchmarks should repeat each request several times per model.
“if you're running this personal benchmark on something that is very critical and there is perhaps a monetary implication or other implications to you making changes to how you work, ideally you should run the same request several times per model”
Open the episode · The Best Way to Test New AI ModelsListen at 21:17
- on AI tool switchingPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Users should change AI tools when benchmark results and practical considerations justify it.
“if all of these tests justify the change, then make the change”
Open the episode · The Best Way to Test New AI ModelsListen at 26:25
- on AI model evaluation criteriaNeutralOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Evaluation criteria and quality standards for AI models change frequently.
“the considerations of what makes you change your mind or what good looks like, those change quite frequently”
Open the episode · The Best Way to Test New AI ModelsListen at 43:11
- on AI model effort settingsNeutralOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Testing multiple effort settings can reveal whether configuration changes evaluation results.
“if you want to be extra diligent, you can always run several models with different effort setting and see if it changes the picture”
Open the episode · The Best Way to Test New AI ModelsListen at 43:59
- on Operational AI tasksNeutralOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Operational AI tasks are likely strongly influenced by the surrounding system setup.
“for operations, I would guess it's going to be most influenced by the overall setup”
Open the episode · The Best Way to Test New AI ModelsListen at 45:11
- on Open-weight modelsPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
For basic tasks, capable open-weight models may perform indistinguishably from commercial models.
“for them, you will probably see that any decent model, let's call it Kimia and above or Mistrel and GLM and above, you will not be able to see any noticeable difference and then open source will be as good as it gets”
Open the episode · The Best Way to Test New AI ModelsListen at 46:11
- on Commercial versus open-weight AI modelsMixedOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Commercial models can retain an advantage on sophisticated tasks.
“For the more aggressive things or more sophisticated stuff, sometimes you will still see the gap between commercial and open source”
Open the episode · The Best Way to Test New AI ModelsListen at 46:23
- on Open-source AI modelsPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Open-source models are currently adequate for most knowledge-work tasks with a good harness.
“as of now, open source is good enough for most knowledge work tasks, especially if the harness in which it's working is a good one”
Open the episode · The Best Way to Test New AI ModelsListen at 46:35
- on Agentic AI toolsPositiveOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Agentic tools require measurable goals rather than vague quality requests.
“Agentic tools need goals. Saying create a good blog post is not measurable.”
Open the episode · The Best Way to Test New AI ModelsListen at 47:16
- on AI subscription plansNegativeOct 7, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
AI subscriptions may become less generous in the future.
“a good prediction to make is that it might not stay as generous in the future”
Open the episode · The Best Way to Test New AI ModelsListen at 48:29
Some agents should remain private while others become team-level agents.
“some agents should stay yours and private while others should become the teams level agents.”
Open the episode · How to Build Team AgentsListen at 4:29
Team agents can reduce knowledge bottlenecks and cross-team coordination problems.
“an agent that is built for the whole team can help with both of these problems”
Open the episode · How to Build Team AgentsListen at 6:12
AI-forward companies have begun merging individual agents into team-level agents.
“They started to merge some of those agents into team-level agents.”
Open the episode · How to Build Team AgentsListen at 7:00
A team agent is one shared agent with common knowledge, memory, and configuration.
“one agent that many people talk to. With shared knowledge, shared memory, and one configuration.”
Open the episode · How to Build Team AgentsListen at 8:23
Team agents handle diverse team tasks, unlike skills focused on specific tasks.
“a skill is a playbook for a specific task, where a team agent is something that your whole team works with on diverse set of tasks.”
Open the episode · How to Build Team AgentsListen at 8:46
- on skill libraries and team agentsNeutralSep 29, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
A skill library and a team agent are distinct systems.
“they are not one and the same.”
Open the episode · How to Build Team AgentsListen at 9:10
Not every agent should be shared.
“not every agent should be shared.”
Open the episode · How to Build Team AgentsListen at 13:42
- on shared knowledge with private agentsPositiveSep 29, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Shared knowledge with private agents is often the easiest starting point.
“the easiest place to start”
Open the episode · How to Build Team AgentsListen at 14:27
Data agents can answer questions spanning multiple company departments.
“it can be the data agent that can answer any data question across multiple departments in the company”
Open the episode · How to Build Team AgentsListen at 15:58
- on private AI agentsPositiveSep 29, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Agents where personal taste outweighs standards should remain private.
“those need to remain private agents”
Open the episode · How to Build Team AgentsListen at 19:18
Unowned or disputed shared knowledge causes team agents to drift rapidly.
“If the team cannot agree on how the work is done or nobody is willing to own and maintain the shared knowledge over time, the agent will drift within weeks, sometimes within days.”
Open the episode · How to Build Team AgentsListen at 19:29
Shared agents drift quickly without knowledge ownership and maintenance.
“the agent will drift within weeks, sometimes within days.”
Open the episode · How to Build Team AgentsListen at 19:36
Teams should delay team agents when complexity exceeds their benefits.
“Don't build it, at least not until you are able to untangle some of the complexities.”
Open the episode · How to Build Team AgentsListen at 20:03
- on Workplace AI agentsPositiveSep 29, 2026· The AI Daily Brief: Artificial Intelligence News and Analysis
Teams should initially limit agents to reading and drafting.
“ideally start with narrower scope, reading and drafting”
Open the episode · How to Build Team AgentsListen at 22:51
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.